The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill suggests an installation method using curl -fsSL https://raw.githubusercontent.com/xdevplatform/xurl/main/install.sh | bash. This pattern is dangerous as it executes unverified code from a third-party repository directly in the shell.
[EXTERNAL_DOWNLOADS]: The skill requires downloading and installing the xurl CLI tool from an external GitHub repository (github.com/xdevplatform/xurl) and via NPM (@xdevplatform/xurl).
[COMMAND_EXECUTION]: The skill is designed to run the xurl CLI utility. It maps user requests to shell commands, which presents a risk of command injection if arguments are not properly handled, although it primarily uses defined subcommands.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection by processing untrusted data from the X API (e.g., via search, mentions, and read).
Ingestion points: Data enters the context via xurl search, xurl mentions, xurl read, and xurl timeline in SKILL.md.
Boundary markers: None identified; the agent is not explicitly told to ignore instructions embedded in the API responses.
Capability inventory: The skill can perform numerous write actions on X (post, reply, delete, DM) and manage application credentials via the xurl CLI.
Sanitization: No sanitization of the retrieved API data is specified.

xurl